Density, Occupancy, and Detectability of Lowland Tapirs, Tapirus Terrestris, in Vale Natural Reserve, Southeastern Brazil

Journal of Mammalogy, 98(1):114–123, 2017 DOI:10.1093/jmammal/gyw118

Density, occupancy, and detectability of lowland tapirs, Tapirus terrestris, in Vale Natural Reserve, southeastern Brazil

Átilla C. Ferreguetti,* Walfrido M. Tomás, and Helena G. Bergallo Department of Ecology, Rio de Janeiro State University, Rua São Francisco Xavier, nº 524, Pavilhão Haroldo Lisboa da Cunha, 2º andar, sala 224. Bairro Maracanã, Rio de Janeiro, RJ CEP 20550-019, Brazil (ACF, HGB) Wildlife Laboratory, Embrapa Pantanal, Rua 21 de Setembro, n° 1.880, Bairro Nossa Senhora de Fátima, Corumbá, Mato Grosso do Sul, CEP 79320-900, Brazil (WMT) * Correspondent: [email protected]

The lowland tapir (Tapirus terrestris, Linnaeus 1758) is one of the surviving members of the Neotropical megafauna. In Brazil, lowland tapirs are considered vulnerable according to the IUCN Red List of Threatened Species and endangered within the Atlantic Rain Forest biome. We aimed to provide the 1st estimates of density and population size for T. terrestris for Vale Natural Reserve (VNR). We predicted the relationships of 6 covariates to habitat occupancy. Density was estimated by the use of distance-sampling techniques, while occupancy, detectability, and activity patterns were assessed with camera-trap monitoring at 39 sample sites over a 1-year period. Density for T. terrestris was 0.8 ± 0.2 lowland tapirs/km2 and population size was 200 ± 33 individuals. Occupancy probability was described by 2 covariates (density of palm trees and distance to water resources) and detectability by those same 2 covariates plus 2 more (distance to road and density of poaching). The species showed the 3 highest peaks of activity at 1900, 2300, and 0400 h. We concluded that VNR still harbors a viable population of lowland tapirs. However, anthropic impacts in the reserve such as poaching and road kills could already be directly affecting the lowland tapir population and producing indirect effects for the whole ecosystem. Results presented herein can be a starting point to support future work in the region and to make predictions regarding the ecosystem relationships, management, and conservation of lowland tapirs.

A anta brasileira (Tapirus terrestris, Linnaeus 1758) é considerada uma das espécies sobreviventes da megafauna neotropical. No Brasil, a anta é considerada vulnerável de acordo com a lista vermelha de espécies ameaçadas da IUCN e considerada em perigo para o bioma da Mata Atlântica. Nosso objetivo foi fornecer as primeiras estimativas de densidade e tamanho populacional de T. terrestris para a Reserva Natural Vale (RNV). Além disso, nós testamos o efeito de 6 covariadas para modelar ocupação do habitat. Estimamos a densidade e o tamanho populacional com a técnica de amostragem de distância por transecção linear. A detectabilidade, ocupação e período de atividade foram estimados com o uso de armadilhas fotográficas em 39 sítios amostrais por um período de 1 ano. Estimamos uma densidade de 0,8 ± 0,2 antas/km2 e um tamanho populacional de 200 ± 33 indivíduos. A probabilidade de ocupação foi representada por 2 covariadas (i.e. densidade de palmeiras e distância do recurso hídrico), e a detectabilidade pelas mesmas covariadas anteriores com a adição de outras 2 (i.e. distância da rodovia e densidade de caça). A população presente na RNV apresentou 3 picos de atividade, às 19:00, 23:00, e 04:00 horas. Nós concluímos que a RNV ainda pode apresentar uma população viável de anta. Porém, os impactos antrópicos presentes na reserva (como a caça ilegal e morte por atropelamentos) podem já estar diretamente afetando a população de anta e produzindo um efeito indireto em todo o ecossistema. Os resultados aqui apresentados podem ser o ponto inicial para o desenvolvimento de futuros estudos na região e também para predições sobre medidas de manejo e conservação da anta e de todo o ecossistema.

Key words: camera trap, distance sampling, poaching, road impact, Tapirus terrestris

114

Downloaded from https://academic.oup.com/jmammal/article-abstract/98/1/114/2977226/Density-occupancy-and-detectability-of-lowland by University of Texas at Austin user on 04 September 2017 FERREGUETTI ET AL.—ECOLOGY OF LOWLAND TAPIRS 115

Data on how species use the habitat are necessary for under- To aid in the conservation of lowland tapirs, our study aimed standing the ecology of each species, including spatial and tem- to provide the 1st estimates of density of the species for the poral distributions (Gentile and Cerqueira 1995; Phillips et al. Vale Natural Reserve (VNR), located in the Atlantic Rain 2004), and are fundamental for management and conservation Forest biome of Brazil. By using occupancy and detectability measures (Caughley 1994; Otis 1997). Each animal species is modeling, we were able to model the spatial distribution of the associated in different ways with structural features of their species, from which we predicted the direction of response of 6 habitat (Downes et al. 1998; Tews et al. 2004); for example, covariates based on prior knowledge of lowland tapir ecology the density of palms (Beck and Terbogh 2002), the presence (Eisenberg and Redford 1999; De Thoisy et al. 2010; Medici of water resources (Goulart et al. 2009), and also elements that 2011; García et al. 2012). A priori, we expected that the covari- can be avoided by a species, for example, poaching and dis- ates measured should represent key habitat features for the spe- tance to the forest edge (Ferreguetti et al. 2015, 2016). Thus, cies (i.e., density of palms, distance to water resources, and tree understanding which factors determine the presence of a spe- density) or elements possibly avoided by the species (i.e., dis- cies in a biome as threatened as the Atlantic Rain Forest biome tance to road, distance to forest edge, and poaching). We tested contributes to the development of conservation measures to the hypothesis that occupancy probability would be higher ensure their survival in such areas (Jones 2001). in sites with higher density of palms, higher density of other Lowland tapirs (Tapirus terrestris, Linnaeus 1758) are a sur- trees, and closeness to water resources. We also hypothesized viving member of the Neotropical megafauna (Emmons and that detectability would be higher in sites with less influence Feer 1997; Hansen and Galetti 2009; De Thoisy et al. 2010) of poaching and greater distance from roads and forest edges. and are distributed throughout South America, except in Chile and Uruguay (Naveda et al. 2008). In Brazil, the lowland tapir is considered vulnerable according to the IUCN Red List of Materials and Methods Threatened Species and considered endangered in the Atlantic Study area.—This study was conducted in the VNR, a pro- Rain Forest biome (Naveda et al. 2008). The lowland tapir is tected area of 235 km2 belonging to the Vale Company. The considered the largest terrestrial mammal occurring in Brazil, reserve is located in the neighboring municipalities of Linhares with low reproductive potential, a long gestation period, and a and Jaguaré (19°06′–19°18′ S and 39°45′–40°19′ W), in north- large home range (Eisenberg and Redford 1999). These biolog- eastern Espírito Santo, Brazil (Fig. 1). ical aspects result in low densities (Medici 2010) and make the The VNR was established through a gradual process of species relatively vulnerable to local extinctions due to demo- land acquisition, which started in 1955, when Vale bought graphic and environmental variations, and losses of genetic its 1st properties in the region, and reached its current extent diversity (Medici et al. 2007). in 1973. The reserve is composed of 1 main block of forest The lowland tapir is still hunted for protein in most of its dis- (approximately 98.1% of the total area), and a much smaller tribution, which represents a major threat to the species (Cullen fragment, known as Biribas Reserve, to the southwest of the et al. 2000; Naveda et al. 2008; Medici 2010). Studies have main block. shown that lowland tapirs select habitat according to 2 main According to the Brazil Vegetation Map (IBGE 1993), the factors: food availability (e.g., species of palms) and water VNR is included within the Atlantic Rain Forest biome, com- resources (Salas and Fuller 1996; Foerster and Vaughan 2002; posed of a mosaic of habitats with 4 main vegetation types Naranjo 2009; Medici 2010). Lowland tapirs have a home (adapted from Jesus 1987; Peixoto and Gentry 1990): “tabu- range of 1.1–14.2 km2 (Medici 2011) and concentrate most of leiro” forest (coastal plain forest), riparian forest, “mussu- their activity in habitat patches (Tobler 2008; Medici 2010). nunga” forest, and natural grassland. The evergreen tabuleiro Lowland tapirs have a low fertility rate, with a single offspring forest has 2 or more upper strata and high densities of lianas resulting after a gestation period of 13–14 months (Barongi and epiphytes and covers approximately 68% of the total area 1993). Because of these characteristics, lowland tapirs have of the reserve. The riparian forest, which covers some 4% of the low abundances and are very susceptible to loss and fragmenta- reserve, is a mixed type of vegetation associated with streams, tion of habitat, poaching, road kills, and diseases transmitted by characterized by widely spaced trees and a predominance of domestic animals (Naveda et al. 2008). palms. The mussununga forest (covering approximately 8% of Although there are good data and information available on low- the VNR) is a type of woody vegetation on sandy soils, physi- land tapirs, some basic knowledge still is lacking in much of its ognomically similar to the tabuleiro forest at an early or inter- distribution. Diet, spatial and temporal distribution, intraspecific mediate stage of regeneration. The natural grasslands occur as interactions, population demography, reproductive parameters, enclaves within the forest, which were once the sites of ponds in and even up-to-date presence/absence data are lacking for many the geological past, and cover ca. 6% of the area of the reserve. localities. These very basic pieces of information are crucial for In addition to these vegetation types, approximately 8% of the modeling the current status, viability, and risk of extinction for reserve is covered by wetlands (swamps) and streams (Fig. 1). lowland tapir populations, as well as predicting the effects of cli- The remaining 6% is composed of administrative structures of mate change and other future scenarios. Good spatial informa- the reserve. tion related to lowland tapirs and their habitats is essential for the Line transect surveys.—To estimate density and abundance development of these models (García et al. 2012). of T. terrestris, we established four 5-km long line transects in

Fig. 1.—Habitat mosaics inside the Vale Natural Reserve, Espírito Santo, Brazil. Black triangles represent the location of each camera trap, and gray lines represent survey transects.

the VNR (Karanth et al. 2004), using the RAPELD protocol single observer starting at sunrise (between 0530 and 0630 h). (Magnusson et al. 2005). It is a technique that seeks to standard- The observer waited 3 h before starting the afternoon survey, ize the collection of biological data and aims to develop meth- between 1300 and 1400 h. The observer walked transects at a ods for long-term ecological research (PELD) and allow rapid speed of approximately 1 km/h. Transects were surveyed twice inventories (Rapid Assessment Program [RAP]). Transects per month with the order being alternated each month. were separated from one other by a minimum of 4 km. For each sighting, we recorded the perpendicular distance of During a 13-month period (April 2013 to May 2014), transects the animal from the transect (measured using measuring tape), were surveyed using distance-sampling techniques (Buckland the length of transect walked to that point, date, timing of the et al. 2001). Each month, transects were surveyed twice by a record, and transect number.

Camera trapping.—We selected 39 sampling sites using a of detection of an animal at a given distance from the track systematic random design stratified by vegetation type to ensure (Laake et al. 1994; Buckland et al. 2001). The best detec- that all 4 of the principal vegetation types found in the VNR tion model was selected by the Akaike Information Criterion were represented (i.e., tabuleiro, riparian, and mussununga (AIC—Akaike 1973). forests, and natural grassland). This scheme was designed to Based on an estimate of home range size of 1.1–14.2 km2 for model occupancy probability of the VNR by the lowland tapir, lowland tapirs (Medici 2011), we checked for spatial and tem- as well as document its activity pattern (Fig. 1). We placed a poral autocorrelation in the records of lowland tapirs for each random grid over a digital map of the reserve and identified the camera trap using Mantel tests (Oksanen et al. 2012). For the sampling points by selecting grid cells. Mantel tests, we calculated a spatial distance matrix using uni- First, in each selected grid cell, we located randomly 1 cam- versal transverse Mercator-14 coordinates (in meters) taken at era trap per grid cell. We did not use bait to attract lowland the center of each grid and a temporal distance matrix using the tapirs. This approach resulted in a relatively even distribution of months of the sampling period. Euclidean distances were used points within the VNR, while maintaining independence among to construct distance matrices for space and time. The Bray– points, which were separated from one another by a distance of Curtis distances were used to a construct distance matrix of the more than 1 km (Magnusson et al. 2005; Ancrenaz et al. 2012). records of lowland tapir per camera. Significance of Mantel At each site, we installed 1 passive infrared Bushnell camera correlations were evaluated by permutation tests with 9,999 trap (Bushnell Outdoor Products, Overland Park, Kansas) in permutations. The analyses were performed in R version 2.15.0 picture function, approximately 40–50 cm above the ground, (R Development Core Team 2012) with the package vegan ver- for continuous surveying throughout the study (from May 2013 sion 2.0-4 for Mantel tests (Oksanen et al. 2012). to June 2014). All stations were examined every 20–25 days to We classified camera-trapping data using 5-day intervals change batteries, when necessary. Traps were programmed to (27 occasions), that, based on the approach of Mackenzie et al. operate for 24 h/day. (2006), would be sufficient to detect or not detect lowland tapirs Covariates.—We used 6 covariates to model occupancy and constructed a reliable detection history. We estimated site probability of the lowland tapir. These were distance to forest occupancy (Ψ) and detection probability (p) for the species, edge (edge_distance), density of trees (with a diameter at breast with 3 possible outcomes: 1) the site was occupied and the spe- height [DBH] of more than 50 cm, tree_density), distance to cies was detected (Ψ × p), 2) the species was present but not water (water_distance), density of palms (palm_density), and detected (Ψ × [1 − p]), and 3) the species was not present and distance to road (road_distance). The potential effect of poach- not detected (1 − Ψ). ing (poaching) on detectability and occupancy also was ana- We estimated detection probabilities by sampling each site lyzed. A priori, these covariates were selected to represent key on 27 occasions. The probability was the parameter projected habitat features (palm_density, tree_density, water_distance) by a maximum likelihood estimation of the proportion of sites or elements possibly avoided by lowland tapirs (road_distance, occupied (Ψ) during the sample period. We used a single- edge_distance). season model. This exercise indicated that the occupancy status At each sampling point, we established 4 plots (30 × 50 m) for tapirs was constant throughout the study, allowing closed arranged by the cardinal compass points (north, south, east, occupancy models to be used (Mackenzie et al. 2006). and west). In each plot, we measured the DBH of each tree In our occupancy analysis, we assessed the covariates that and counted the number of large trees (DBH > 50 cm). In these might affect occupancy and detectability, in an attempt to iden- plots, we also counted the number of individuals of each spe- tify habitat preferences. We constructed a set of 20 candidate cies of palms. models selected by a priori hypotheses based on 4 different Three spatial covariates—distance to forest edge, water approaches: 1) considering occupancy probability and detect- resource, and main road—were quantified for each of the 39 ability as constant across all sites, 2) considering the variation sample sites using ArcGIS software (ESRI*ArcMap 10.1, in occupancy as a function of covariates, 3) considering the Redlands, California—ESRI 2011). variation in detectability as a function of covariates, and 4) con- Density of poaching within the study area was calculated sidering both the variation in occupancy and detectability as a using the georeferenced database of a 10-year period in which function of covariates. This allowed us to evaluate the differ- poaching events were recorded by the reserve’s security guards ences in habitat occupancy as determined by a single covariate (source: VNR). The density of poaching (records per km2) was or a set of covariates, which would contribute to an improve- calculated for each grid cell in which a camera trap had been ment in the model’s performance. installed by dividing the number of records by the grid cell Occupancy modeling was run in PRESENCE 9.3 software area. (Mackenzie and Royle 2005) with 2,000 bootstraps to assess Data analysis.—Density and population size were esti- the adjustment fit (P) and overdispersion parameter (ĉ). In our mated by total number of individuals observed along each trail assessment of occupancy closure and the factors that influ- through the program DISTANCE 6.2 (Buckland et al. 2001). enced occupancy and detection, we ranked the models by AIC DISTANCE uses the perpendicular distances (animal-track) adjusted for small sample size (AICc), following Burnham and to estimate effective strip width (ESW) in the study area and Anderson (2002). All models with a ΔAICc value < 2 were model the detection function that best suits the probability considered to be competitive. We also used the AICc weight

(wi) for each model, which corresponds to the amount of evi- the study. We obtained 2,134 independent detections of T. ter- dence in favor of a given model, to choose the best model that restris and observed the species in 30 of the 39 sites, which we used to test our hypotheses. resulted in a naive occupancy of 0.77 and a detectability of The circadian cycle of the species was described based on 0.52. The northern portion of VNR resulted in the highest the timing of photographs obtained from the camera traps. To occupancy rates and detectability of lowland tapirs (Ψ = 0.58– avoid pseudoreplication, records were only considered to be 0.95 and p = 0.31–0.48), followed by the southern portion independent when separated by an interval of 24 h. Analyses (Ψ = 0.32–0.43 and p = 0.18–0.29). The west portion resulted were conducted in the “Circular” package in R with associated in the lowest occupancy rates and detectability (Ψ = 0.08–0.13 analytic packages. Circular summaries (Lund and Agostinelli and p = 0.06–0.09). 2007) were used to determine the mean overall timing of activ- From the 20 occupancy models produced (Table 1), occu- ity as recorded by camera traps. A resampling approach was pancy was best described by 2 covariates: 1) density of palm used to derive unbiased estimates of the 95% confidence inter- (palm_density), which had a positive relationship in which vals, whereby the circular mean from a random sample of 100 occupancy increased with high palm density (Ψ = 0.11–0.98; data points was calculated 10,000 times with replacement to Fig. 2A) and 2) distance to water resources (water_distance), generate reliable estimates. which had a negative relationship, in which occupancy by lowland tapirs decreased as the distance to water increased (Ψ = 0.009–0.96; Fig. 2C). Detectability was affected by 4 Results covariates: 1) palm_density, which had a positive relationship Density and population size.—A total of 908 km of transects in which detectability increased in sites with higher density of (196 samples) were surveyed in this study. Throughout the sur- palms (p = 0.11–0.48; Fig. 2B); 2) water_distance, which had veys, T. terrestris was sighted 63 times. a negative relationship with higher detectability in sites with Estimated density for T. terrestris was 0.8 ± 0.2 lowland shorter distances to water (p = 0.02–0.43; Fig. 2D); 3) Density tapirs/km2 and estimated population size was 200 ± 33 individu- of poaching (poaching), which had a negative relationship als with an ESW of 10.80 ± 1.12 m with observations obtained illustrating that the higher the poaching the lower detectability from 0 to 36 m from the line of the transect. The model that best (p = 0.16–0.41; Fig. 2E); and (4) distance to road (road_dis- fitted our data was a Half Normal key with cosine adjustment. tance), which had a negative relationship with higher detect- The coefficient of variation for both parameters was 16.6%. ability in sites far away from roads (p = 0.1–0.4; Fig. 2F). Occupancy and detectability models.—Records of the Activity pattern.— Tapirus terrestris showed a broad activity lowland tapir were not autocorrelated with respect to space pattern, being photographed throughout the 24-h cycle (Fig. 3). (Mantel’s r = −0.02, P = 0.65) or time (Mantel’s r = 0.05, The species showed 3 higher peaks of activity at 1900, 2300, P = 0.23). A total of 7,020 trap-days were conducted during and 0400 h.

Table 1.—Occupancy models for Tapirus terrestris in the Vale Natural Reserve, Brazil, estimated by camera trapping between May 2013 and June 2014, grouped in sampling intervals of 5 consecutive days. With (.) = as constant and covariates: distance to forest edge (edge_dist); density of trees with diameter breast height > 50 cm (tree_dens); density of poaching (poaching); density of palms (palm_dens); distance to road (road_dist); and distance to water resources (water_dist). Ψ = occupancy; p = detectability; overdispersion parameter (ĉ) = 1.14 and the adjustment fit (P) = 0.23.

Model AICc ΔAICc wi Number of parameters Ψ(palm_dens;water_dist);p(poaching;water_dist;road_dist;palm_dens) 532.62 0 0.89 8 Ψ(palm_dens;water_dist);p(poaching;road_dist;palm_dens) 540.27 7.65 0.08 7 Ψ(palm_dens;water_dist);p(poaching;water_dist;road_dist) 541.97 9.35 0.001 7 Ψ(palm_dens;water_dist);p(poaching;palm_dens) 542.01 9.39 < 0.001 6 Ψ(palm_dens);p(poaching;water_dist) 542.33 9.71 < 0.001 5 Ψ(palm_dens;water_dist);p(road_dist;palm_dens) 556.10 23.48 < 0.001 6 Ψ(palm_dens);p(road_dist;poaching) 556.50 23.88 < 0.001 5 Ψ(water_dist);p(road_dist;poaching) 561.14 28.52 < 0.001 5 Ψ(water_dist);p(poaching) 565.85 33.23 < 0.001 4 Ψ(palm_dens);p(poaching) 572.78 40.16 < 0.001 4 Ψ(.);p(.) 576.70 44.08 < 0.001 2 Ψ(.);p(road_dist;poaching) 576.84 44.18 < 0.001 3 Ψ(palm_dist);p(.) 577.03 44.41 < 0.001 3 Ψ(water_dist);p(.) 577.78 45.16 < 0.001 3 Ψ(.);p(poaching) 577.88 45.26 < 0.001 3 Ψ(.);p(dist_road) 577.98 45.36 < 0.001 3 Ψ(poaching;edge_dist);p(poaching) 578.40 45.78 < 0.001 5 Ψ(road_dist);p(.) 578.65 46.03 < 0.001 3 Ψ(poaching);p(.) 578.92 46.30 < 0.001 3 Ψ(edge_dist);p(.) 579.02 46.40 < 0.001 3

Fig. 2.—A) Relationship between occupancy of the Atlantic Rain Forest biome, Vale Natural Reserve, Espírito Santo, Brazil, by Tapirus terrestris and density of palms. B) Relationship between detectability of T. terrestris and density of palms. C) Relationship between occupancy of the Atlantic Rain Forest biome, Vale Natural Reserve, Espírito Santo, Brazil, by T. terrestris and distance to water resources. D) Relationship between detectability of T. terrestris and distance to water resources. E) Relationship between detectability of T. terrestris in the Atlantic Rain Forest biome, Vale Natural Reserve, Espírito Santo, Brazil, and density of poaching. F) Relationship between detectability of T. terrestris and distance to road as estimated by camera trapping between May 2013 and June 2014, grouped in sampling intervals of 5 consecutive days.

Discussion value recommended to obtain reliable estimates (Buckland et al. 2001). Thus, our density estimate serves as a starting Density and population size.—The density of lowland tapirs point for monitoring the lowland tapir in the VNR and for com- reported in this study, estimated using line transects in the parison with other studies that have used the same methodol- VNR, is the 1st estimate of density for this area and showed ogy (Table 2). Density estimates for the species in the Atlantic a coefficient of variation below 20%, which is the maximum

Forest were lower than those reported in the present study. This lowland tapirs, which may be considered a viable population is because those estimates were made in the fragmented areas for conservation (Gatti et al. 2011). of Morro do Diabo (0.34–0.64 individuals per km2—Medici Occupancy and detectability models.—The model to pre- 2010) and the Ecologic estation of Caetetus (0.47 individuals dict potential habitat suitability for lowland tapirs in the VNR per km2—Cullen et al. 2000), both in the state of Sao Paulo. showed that among the 6 covariates evaluated in this study, 2 However, our estimate is similar to those obtained in more best explained habitat occupancy by lowland tapirs (i.e., distance preserved areas in other biomes such as Pantanal (0.30–1.01 to water resources and density of palms) and 4 were best able to individuals per km2—Trolle et al. 2008) and Amazonian Forest predict detectability for the species (i.e., density of palms, dis- (3.3–3.7 individuals per km2—Mendes-Pontes 2004). Those tance to water resources, poaching, and distance to road). density estimates were obtained using different methodologies, The relationship between occupancy by lowland tapirs and which makes comparisons between studies difficult. However, distance to water resources was expected after it was deter- herein we show that VNR contains a population of 200 ± 33 mined which habitats they usually occupied, such as riparian forests, wetlands, and streams, where they are most active, especially when foraging (Medici 2010), making it easy to detect individuals in sites closest to a water resource. Thus, the highest occupancy rates and detectability of the lowland tapir in the northern portion of the VNR can be attributed to this portion of the reserve having a higher quantity of water resources (Kierulff et al. 2015) compared to the other portions of VNR (i.e., south and west). Norris (2014) found that distance to water was one of the covariates that best described the distribution of lowland tapirs. Lowland tapirs also used sites closest to water as a place for resting, defecation, to prevent ectopara- sites, to facilitate their movements between foraging sites, and for cooling off during the hottest hours of the day (Padilla and Dowler 1994; Foerster and Vaughan 2002). Brazil currently is experiencing a hydro crisis that began in 2014. Lack of rainfall (CPTEC 2015), along with the creation of artificial barrages to store water, has damaged ecosystems and animal populations, especially species like the lowland tapir which is highly depen- dent on water (García et al. 2012). Many streams are disappear- Fig. 3.—Activity pattern of T. terrestris in the Atlantic Rain Forest, ing because of this process. Vale Natural Reserve, Espírito Santo, Brazil, as estimated by camera Along with water resources, the density of palms is also trapping between May 2013 and June 2014. Inside values 0–30 are the a covariate that represents a habitat key to the lowland tapir, frequency of records and outside values are hours of the day. because the fruits produced by these palm trees are crucial food

Table 2.—Density estimates of Tapirus terrestris per km2 in the Brazilian biomes. FIT = Footprint Identification Technique; ind = individuals.

Biome Location Density (ind/km2) Reference and applied method Atlantic Forest Ecologic Estation Caetetús, São Paulo 0.47 Cullen et al. (2000), line transect Atlantic Forest State Park Morro do Diabo, São Paulo 0.20–0.84 Cullen et al. (2000), line transect Atlantic Forest Fazenda Mosquito, São Paulo 0.3 Cullen et al. (2000), line transect Atlantic Forest State Park Mata dos Godoy, Paraná 2.20–2.50 Rocha (2001), number of observations/area Atlantic Forest State Park Morro do Diabo, São Paulo 0.43 Medici (2010), FIT Atlantic Forest State Park Morro do Diabo, São Paulo 0.64 Medici (2010), line transect Atlantic Forest State Park Morro do Diabo, São Paulo 0.34 Medici (2010), radiotelemetry Atlantic Forest Caraguatatuba, São Paulo 0.13–0.21 Ramírez (2013), genetic analysis Atlantic Forest Intervales, São Paulo 0.20–0.57 Ramírez (2013), genetic analysis Atlantic Forest Vale Natural Reserve, Espírito Santo 0.6–1.0 Present study, line transect Amazon Forest Brazilian Amazon 0.11–0.52 Peres (2000), line transect Amazon Forest Center of ecological studies Kayapó, Para 0.62 Zimmerman et al. (2001), line transect Amazon Forest Ecological Estation of Maracá, Roraima 3.3–3.7 Mendes-Pontes (2004), line transect Amazon Forest Lake Uauaçú, Mid-West Amazon 0.23 Haugaasen and Peres (2005), line transect Pantanal Fazenda Acurizal, Paraguay River 0.64 Schaller (1983), direct counts Pantanal Fazenda Nhumirim, Nhecolândia Forest: 0.40; Cerrado: 0.13 Desbiez (2007) and Desbiez (2009), line transect Pantanal SESC Pantanal, Barão de Melgaço Forest: 0.71; open areas: 0.37 Cordeiro (2004), line transect Pantanal SESC Pantanal, Barão de Melgaço 0.47–0.69 Trolle et al. (2008), camera trap Pantanal SESC Pantanal, Barão de Melgaço 0.30–1.01 Trolle et al. (2008), line transect

resources for lowland tapirs (Fragoso 1997). The northern part For example, in Alberta, Canada, the construction of these of the reserve had the highest occupancy rates and palm den- structures favored the passage of large mammals such as bears sity values, which were expected as much of the riparian for- (Ursus arctos), wolves (Canis lupus), moose (Cervus elaphus), ests (i.e., vegetation type that presents high dominance of palm and deer (Odocoileus sp.) along the Trans-Canada highway trees) are located in the northern VNR (Kierulff et al. 2015). (Clevenger and Waltho 2005). In the Atlantic Rain Forest biome of Brazil, lowland tapirs are Activity period.—Lowland tapirs have been described as known to use patches of Jerivá palm (Syagrus romanzoffiana) nocturnal or crepuscular (Padilla and Dowler 1994; Noss very intensively (Galetti et al. 2001; Tófoli 2006). In the north- et al. 2003; Tobler 2008; Medici 2010; Wallace et al. 2012; eastern region of the Brazilian Pantanal, lowland tapirs show Cruz et al. 2014); however, several factors can affect the activ- a high preference for Acuri palm forests (Attalea phalerata) ity period of this species (i.e., poaching, rainy season). For low- compared to other vegetation types (Cordeiro 2004). In the land tapirs, increased diurnal activity was observed in humid Amazon, patches of Mauritia flexuosa frequently are visited months (Foerster and Vaughan 2002; Medici 2010). Lowland by lowland tapirs (Bodmer 1990; Tobler 2008). Accordingly, tapirs usually are inactive in the middle of the day, which cor- occupancy and detectability models demonstrate high occur- responds to the hottest hours of the day (Medici 2010). From rence probabilities in such habitat areas. personal observation, we noticed that during the hottest hours Besides natural habitat factors, anthropic factors (i.e., poach- of the line transect survey, lowland tapirs were found bedded ing and roads) potentially could influence the detectability of down in the forest interior. lowland tapirs (Licona et al. 2011). Poaching was reported as a Implications for conservation.—Although the present study problem for animal populations in VNR since 1999 (Chiarello was conducted in the VNR, the northern region of the state of 2000) and the west portion of the VNR is the most affected by Espírito Santo holds a connected block of protected area of 550 poaching (Kierulff et al. 2015). The high rate of poaching in km2. This protected block has 2 important purposes: 1) moni- the western portion of the VNR was associated with the lowest toring the lowland tapir population in the state of Espírito Santo, occupancy rates and detectability of lowland tapirs. The detect- and 2) quantifying the impact of roads on animal populations. ability of lowland tapirs has likely been affected by poaching Results presented herein can be a starting point to support as it is known that tapirs tend to avoid areas with higher hunt- future work in the region and to make predictions regarding ing activity (Laundre et al. 2010). Medici et al. (2007) con- the ecosystem relationships, management, and conservation of ducted a population and habitat viability analysis for lowland lowland tapirs. Furthermore, the results can be used as a sur- tapirs and found that poaching is one of the major threats to the rogate for other regions or biomes in which the species occurs, species, which is consistent with previous research where the because many locations may be affected by the same covariates presence of lowland tapirs and the level of poaching were nega- used herein (i.e., poaching and road impact). tively correlated (Peres 2000; Cullen et al. 2001; Fa et al. 2002; Cruz et al. 2014). Lowland tapirs especially are vulnerable to hunting when compared with other ungulates (e.g., cervids— Acknowledgments Ferreguetti et al. 2015) because of their relatively low-density We thank Vale Natural Reserve for their support of this research. population (Bodmer et al. 1997; Medici et al. 2007) and life We appreciate the considerations of both anonymous reviewers. history traits (i.e., long-lived individuals, low rates of increase, AFC is also grateful to Coordenacão de Aperfeiçoamento de and long generation times—Bodmer et al. 1997; Emmons and Pessoal de Nível Superior (CAPES) for a graduate scholarship. Feer 1997). Poaching not only has a negative effect on lowland HGB thanks FAPERJ (E-26/103.016/2011), Prociência/UERJ, tapir populations, but also has an indirect effect on the compo- and CNPq (457458/2012-7, 307715/2009-4) for research and sition and structure of the forest, because lowland tapirs play productivity grants. This study is portion of the results of the a critical ecological role, affecting the structure, composition, Programa de Pesquisas em Biodiversidade da Mata Atlântica growth, and regeneration of vegetation (Bodmer 1990, 1991; (PPBioMA) which was supported by Conselho Nacional de Dirzo and Miranda 1991; Fragoso 1997; Galetti et al. 2015). Desenvolvimento Científico e Tecnológico (CNPq) (Process Detectability was lowest in sites closest to the road that cuts No. 457458/212-7). Data storage is supported by PPBio MA the VNR, suggesting that lowland tapirs could be avoiding (CNPq - 457458/2012-7) and FAPERJ (E-26/111.394/2012). areas next to roads. Roads and other human developments may act as barriers, because the animals tend to avoid sites closest Literature Cited to roads, a behavior that has been reported for other large mammals occurring in contiguous habitats (Paetkau et al. 1997). Akaike, H. 1973. Information theory and an extension of the maximum likelihood principle. Pp. 267–281 in Second international Combined with our finding that detectability is lower in sites symposium on information theory (B. N. Petrov and F. Csaki, closest to the road, during the period of June 2014 to June 2015, eds.). Akademiai, Budapest, Hungary. 4 individuals of lowland tapir were found dead in the road that Ancrenaz, M., A. Hearn, J. Ross, R. Sollmann, and A. Wilting. cuts the reserve (i.e., BR 101 road), and 1 of the individuals 2012. Handbook for wildlife monitoring using camera-traps. Small was a pregnant female. To maintain the flow of lowland tapirs Carnivore Specialist Group. BBEC II Secretariat, Sabah, Malaysia. across roads, construction of special structures such as wildlife Barongi, R. A. 1993. Husbandry and conservation of tapirs Tapirus crossings have been proposed (Foster and Humphrey 1995). spp. International Zoo Yearbook 32:7–15.

Beck, H., and J. Terbogh. 2002. Groves versus isolates: how spatial Eisenberg, J., and K. H. Redford. 1999. Mammals of the Neotropics. aggregation of Astrocaryum murumuru palms affects seed removal. University of Chicago Press, Chicago, Illinois 3:1–624. Journal of Tropical Ecology 18:275–288. Emmons, L. H., and F. Feer. 1997. Neotropical rainforest mammals: a Bodmer, R. 1990. Fruit patch size and frugivory in the lowland tapir field guide. 2nd ed. University of Chicago Press, Chicago, Illinois. (Tapirus terrestris). Journal of Zoology (London) 222:121–128. ESRI. 2011. ArcGIS desktop: release 10. Environmental Systems Bodmer, R. E. 1991. Strategies of seed dispersal and seed predation Research Institute, Inc. Redlands, California. in Amazonian ungulates. Biotropica 23:255–261. Fa, J. E., C. A. Peres, and J. A. Meeuwig. 2002. Bushmeat exploration Bodmer, R. E., J. F. Eisenberg, and K. H. Redford. 1997. Hunting and in tropical forests: an intercontinental comparison. Conservation the likelihood of extinction of Amazonian mammals. Conservation Biology 16:232–237. Biology 11:460–466. Ferreguetti, Á. C., W. M. Tomás, and H. G. Bergallo. 2015. Buckland, S. T., D. R. Anderson, K. P. Burnham, J. L. Laake, D. Density, occupancy, and activity pattern of two sympatric deer L. Borchers, and L. Thomas. 2001. Introduction to distance sam- (Mazama) in the Atlantic Forest, Brazil. Journal of Mammalogy pling: estimating abundance of biological populations. 1st ed. 96:1245–1254. Oxford University Press, Oxford, United Kingdom. Ferreguetti, A. C., W. M. Tomas, and H. G. Bergallo. 2016. Burnham, K., and D. Anderson. 2002. Model selection and multi- Density and niche segregation of two armadillo species (Xenarthra: model inference: a practical information and theoretic approach. Dasypodidae) in the Vale Natural Reserve, Brazil. Mammalian 2nd ed. Springer. Biology 81:138–145. Caughley, G. 1994. Directions in conservation biology. Journal of Foerster, C. R., and C. Vaughan. 2002. Home range, habitat use, and Animal Ecology 63:215–244. activity of Baird’s tapir. Biotropica 34:423–437. Chiarello, A. G. 2000. Influência da caça ilegal sobre mamíferos e Foster, M. L., and S. R. Humphrey. 1995. Use of highway under- aves das matas de tabuleiro do norte do estado do Espírito Santo. passes by Florida panthers and other wildlife. Wildlife Society Boletim do Museu de Biologia Mello Leitão 12:229–247. Bulletin 23:95–100. Clevenger, A. P., and N. Waltho. 2005. Performance indices to iden- Fragoso, J. M. V. 1997. Tapir-generated seed shadows: scale-depen- tify attributes of highway crossing structures facilitating movement dent patchiness in the Amazon rain forest. Journal of Ecology of large mammals. Biological Conservation 121:453–464. 85:519–529. Cordeiro, J. L. P. 2004. Estrutura e heterogeneidade da paisagem Galetti, M., R. S. Bovendorp, and R. Guevara. 2015. Defaunation de uma unidade de conservação no nordeste do pantanal (RPPN of large mammals leads to an increase in seed predation in the SESC Pantanal), Mato Grosso, Brasil: efeitos sobre a distribuição Atlantic forests. Global Ecology and Conservation 3:824–830. e densidade de antas (Tapirus terrestris) e de cervos-do-Pantanal Galetti, M., A. Keuroghlian, L. Hanada, and M. I. Morato. 2001. (Blastocerus dichotomus). Ph.D. thesis, Rio Grande do Sul Federal Frugivory and seed dispersal by the lowland tapir (Tapirus terres- University, Rio Grande do Sul, Brazil. tris) in Southeast Brazil. Biotropica 33:723–726. CPTEC. 2015. Centro de Previsão de Tempo e Estudos Climáticos. García, M. J., E. P. Medici, E. J. Naranjo, W. Novarino, and R. http://www.cptec.inpe.br/. Accessed 1 November 2015. S. Leonardo. 2012. Distribution, habitat and adaptability of the Cruz, P., A. Paviolo, R. F. Bó, J. J. Thompson, and M. S. Di Bitetti. genus Tapirus. Integrative Zoology 7:346–355. 2014. Daily activity patterns and habitat use of the lowland tapir Gatti, A., D. Brito, and S. L. Mendes. 2011. How many lowland (Tapirus terrestris) in the Atlantic Forest. Mammalian Biology tapirs (Tapirus terrestris) are needed in Atlantic Forest fragments 79:376–383. to ensure long-term persistence? Studies on Neotropical Fauna and Cullen, L., R. E. Bodmer, and C. Valladares Pádua. 2000. Effects Environment 46:77–84. of hunting in habitat fragments of the Atlantic forests, Brazil. Gentile, R., and R. Cerqueira. 1995. Movement patterns of five spe- Biological Conservation 95:49–56. cies of small mammals in a Brazilian restinga. Journal of Tropical Cullen, L., R. E. Bodmer, and C. Valladares-Padua. 2001. Ecology 11:671–677. Ecological consequences of hunting in Atlantic forest patches, Sao Goulart, F. V. B., N. C. Cáceres, M. E. Graipel, M. A. Tortato, I. Paulo, Brazil. Oryx 35:137–144. R. Ghizoni, and L. G. R. Oliveira-Santos. 2009. Habitat selec- Desbiez, A. L. J. 2007. Wildlife conservation in the Pantanal: habitat tion by large mammals in a southern Brazilian Atlantic Forest. alteration, invasive species and bushmeat hunting. Ph.D. disserta- Mammalian Biology 74:182–190. tion, University of Kent, Canterbury, United Kingdom. Hansen, D. M., and M. Galetti. 2009. The forgotten megafauna. Desbiez, A. L. J. 2009. Lowland tapirs in the Nhecolânia Region of Science 324:42–43. Brazilian Pantanal: population density, habitat use and threats. Haugaasen, T., & C. A. Peres. 2005. Mammal assemblage structure Tapir Conservation 18:7–10. in Amazonian flooded and unflooded forests. Journal of Tropical De Thoisy, B., et al. 2010. Population history, phylogeography, and Ecology 21:133–145. conservation genetics of the last Neotropical mega-herbivore, the IBGE. 1993. Mapa de Vegetação do Brasil. Fundação Instituto lowland tapir (Tapirus terrestris). BMC Evolutionary Biology Brasileiro de Geografia e Estatística, Rio de Janeiro, Brazil. 10:1. Jesus, R. M. 1987. Mata Atlântica de Linhares: Asectos Florestais. Dirzo, R., and A. Miranda. 1991. Altered patterns of herbivory and pp. 35–53 in Desenvolvimento Econômico e Impacto Ambiental diversity in the forest understory: a case study of the possible con- em Áreas de Trópico Úmido Brasileiro – A experiência da CVRD sequences of contemporary defaunation. Pp. 273–287 in Plant ani- (Secretaria Estadual de Meio Ambiente, Idaho Water Resource mal interactions: evolutionary ecology in tropical and temperate Board, and Companhia Vale do Rio Doce, eds.). Companhia Vale regions (P. W. Price, T. M. Lewinsohn, G. W. Fernández, and W. W. do Rio de Doce, Rio de Janeiro, Brazil. Benson, eds.). Wiley Press, New York. Jones, J. 2001. Habitat selection studies in avian ecology: a critical Downes, B. J., P. S. Lake, E. S. G. Schreiber, and A. Glaister. 1998. review. Auk 118:557–562. Habitat structure and regulation of local species diversity in a stony Karanth, K. U., J. D. Nichols, N. S. Kumar, W. A. Link, and J. upland stream. Ecological Monographs 68:237–257. E. Hines. 2004. Tigers and their prey: predicting carnivore densities

from prey abundance. Proceedings of the National Academy of Otis, D. L. 1997. Analysis of habitat selection studies with multi- Sciences 101:4854–4858. ple patches within cover types. Journal of Wildlife Management Kierulff, M. C. M., L. H. S. Avelar, M. E. S. Ferreira, K. F. Povoa, 61:1016–1022. and R. S. Bernils. 2015. Reserva Natural Vale: história e aspectos Padilla, M., and C. Dowler. 1994. Tapirus terrestris. Mammalian físicos. Ciência & Ambiente 49:7–40. Species 481:1–8. Laake, J., S. Buckland, D. Anderson, and K. Burnham. 1994. Paetkau, D., L. P. Waits, P. L. Clarkson, L. Craighead, and Distance user’s guide. Colorado State University Press, Fort C. Strobeck. 1997. An empirical evaluation of genetic distance sta- Collins. tistics using microsatellite data from bear (Ursidae) populations. Laundre, J. W., L. Hernandez, and W. J. Ripple. 2010. The landscape Genetics 147:1943–1957. of fear: ecological implications of being afraid. The Open Ecology Peixoto, A., and A. Gentry. 1990. Diversity and floristic composition Journal 3:1–7. of the “Mata de Tabuleiro” of the Linhares Forest Reserve, Espirito Licona, M., R. Mccleery, B. Collier, D. J. Brightsmith, and Santo, Brazil. Revista Brasileira de Botânica 13:19–25. R. Lopez. 2011. Using ungulate occurrence to evaluate commu- Peres, C. A. 2000. Effects of subsistence hunting on vertebrate com- nity-based conservation within a biosphere reserve model. Animal munity structure in Amazonian forests. Conservation Biology Conservation 14:206–214. 14:240–253. Lund, U., and C. Agostinelli. 2007. Circular: circular statistics. Phillips, M. L., W. R. Clark, S. M. Nusser, M. A. Sovada, and R. http://cran.r-project.org/package5circular. Accessed 30 June 2015. J. Greenwood. 2004. Analysis of predator movement in prairie Mackenzie, D., J. Nichols, J. Royle, K. Pollock, L. Bailey, and landscapes with contrasting grassland composition. Journal of J. Hines. 2006. Occupancy estimation and modeling: inferring pat- Mammalogy 85:187–195. terns and dynamics of species occurrence. Elsevier Publishing, Ramírez, J. F. M. 2013. Diversidade genética e estrutura populacio- New York. nal da anta (Tapirus terrestris) na Serra do Mar, São Paulo, Brasil. Mackenzie, D. I., and J. A. Royle. 2005. Designing occupancy stud- Master’s thesis, Universidade Estadual Paulista Júlio de Mesquita ies: general advice and allocating survey effort. Journal of Applied Filho, Instituto de Biociências de Rio Claro, São Paulo, Brazil. Ecology 42:1105–1114. R Development Core Team. 2012. R: a language and environment Magnusson, W. E., et al. 2005. RAPELD: Uma modificação do for statistical computing. R Foundation for Statistical Computing, método de Gentry para o levantamento da biodiversidade em áreas. Vienna, Austria. http://R-project.org/. Accessed 10 March 2016. Biota Neotropica 5. http://www.biotaneotropica.org.br/v5n2/pt/ Rocha, V. J. 2001. Ecologia de mamíferos de médio e grande porte do abstract?point-of-view+bn0100522005. Accessed 15 October Parque Estadual Mata dos Godoy, Londrina (PR). Ph.D. disserta- 2014. tion, Universidade Federal do Paraná, Paraná, Brazil. Medici, E. P., et al. 2007. Lowland tapir (Tapirus terrestris) pop- Salas, L. A., and T. K. Fuller. 1996. Diet of the lowland tapir ulation and habitat viability assessment (PHVA): final report. (Tapirus terrestris) in the Tabaro River valley, southern Venezuela. Relatório Técnico. IUCN/SSC Tapir Specialist Group (TSG) and Canadian Journal of Zoology 74:1444–1451. IUCN/SSC Conservation Breeding Specialist Group (CBSG). Schaller, G. B. 1983. Mammals and their biomass in a Brazilian Medici, E. P. 2010. Assessing the viability of lowland tapir popula- ranch. Arquivos de Zoologia do Estado de São Paulo 31:1–36. tions in a fragmented landscape. Ph.D. thesis, Kent University, Tews, J.,et al. 2004. Animal species diversity driven by habitat het- London, United Kingdom. 276 p. erogeneity/diversity: the importance of keystone structures. Journal Medici, E. P. 2011. Family Tapiridae (tapirs). Pp. 24–56 in Handbook of Biogeography 31:79–92. of the mammals of the world, volume 2: hoofed mammals first Tobler, M. W. 2008. The ecology of the lowland tapir in Madre de (D. Wilson and R. Mittermeier, eds.). Lynx Editions, Barcelona, Dios, Peru: using new technologies to study large rainforest mam- Spain. mals. Ph.D. dissertation, Texas A&M University, College Station. Mendes-Pontes, A. R. 2004. Ecology of a community of mammals in Tófoli, C. F. D. 2006. Frugivoria e dispersão de sementes por Tapirus a seasonailly dry forest in Roraima, Brazilian Amazon. Mammalian terrestris (Linnaeus, 1758) na paisagem fragmentada do Pontal Biology 69:319–336. do Paranapanema, São Paulo. Ph.D. thesis, Universidade de São Naranjo, E. J. 2009. Ecology and conservation of Baird’s tapir in Paulo, São Paulo, Brazil. Mexico. Tropical Conservation Science 2:140–158. Trolle, M., A. J. Noss, J. L. P. Cordeiro, and L. F. B. Oliveira. 2008. Naveda, A., et al. 2008. Tapirus terrestris. www.iucnredlist.org. Brazilian tapir density in the Pantanal: a comparison of systematic Accessed 1 June 2015. camera-trapping and line-transect surveys. Biotropica 40:211–217. Norris, D. 2014. Model thresholds are more important than pres- Wallace, R., G. Ayala, and M. Viscarra. 2012. Lowland tapir ence location type: understanding the distribution of lowland tapir (Tapirus terrestris) distribution, activity patterns and relative abun- (Tapirus terrestris) in a continuous Atlantic forest of Southeast dance in the Greater Madidi-Tambopata Landscape. Integrative Brazil. Tropical Conservation Science 7:529–547. Zoology 7:407–419. Noss, A. J., et al. 2003. A camera trapping and radio telemetry Zimmerman, J. A., C. L. Kelchner, P. A. Klein, J. C. Hamilton, & S. study of lowland tapir (Tapirus terrestris) in Bolivian Dry M. Foiles. 2001. Surface step effects on nanoindentation. Physics Forests. Newsletter of the IUCN/SSC Tapir Specialist Group Review Letters 87:165507–165510. 12:24–32. Oksanen, J., et al. 2012. Vegan: community ecology package. R Submitted 18 February 2016. Accepted 22 June 2016. package version 2.0-4. http://CRAN.R-project.org/package=vegan. Accessed 10 March 2016. Associate Editor was Leslie N. Carraway.

Downloaded from https://academic.oup.com/jmammal/article-abstract/98/1/114/2977226/Density-occupancy-and-detectability-of-lowland by University of Texas at Austin user on 04 September 2017